Staff ML Engineer – Post-Training & Sandbox Systems

Dex; Greater London; Hybrid; GBP 90,000 - 130,000; Full time

Job summary

Dex, partnered with frontier AI labs, seeks a Member of Technical Staff to own the post-training and harness loop, bridging research and production. You design post-training pipelines, sandboxed environments, and data infrastructure to train agents on complex tasks.

Based in London or Paris, you will lead SFT/RL workflows, build reproducible evaluation frameworks, and collaborate with researchers and software engineers to deliver production-grade systems.

Qualifications

• Experience directly training or fine-tuning models using SFT, RL, or RLHF.
• Strong foundations in machine learning concepts, optimization, RL architectures and evaluation.
• Production-grade Python engineering with modular, observable systems.
• Ability to take ambiguous, first-principles problems from concept to working system over multi-month horizons.

Responsibilities

• Design and execute end-to-end post-training pipelines using SFT and RL.
• Architect sandboxed execution environments and deterministic feedback loops for safe interaction with real-world software.
• Build and scale data infrastructure converting workflows into curated datasets for benchmarks.
• Establish evaluation frameworks to benchmark agents and ensure reproducible iterations.
• Provide technical leadership translating research goals into production-grade systems.

Skills

SFT/RL training

Python programming

ML fundamentals

Production-grade systems

Tools

Python

SFT

RL

Job description

Dex, partnered with frontier AI labs, seeks a Member of Technical Staff to own the post-training and harness loop, bridging research and production. You design post-training pipelines, sandboxed environments, and data infrastructure to train agents on complex tasks.

Based in London or Paris, you will lead SFT/RL workflows, build reproducible evaluation frameworks, and collaborate with researchers and software engineers to deliver production-grade systems.